Table 3.

Experiments using the databases CIFAR-100 and MNIST, taking the error percentage as evaluation metric. We only indicate the best results with bold.

Algoritmo	CIFAR-100	MNIST
Deep Co-Training	38.77	—–
Co-Teaching	38.13	8.08
DSSCo-Training	37.21	7.10
Classical Co-Training	61.02	34.86